Fifty-one formalin-fixed, paraffin-embedded osteosarcoma specimens (before the administration of neo-adjuvant chemotherapy) were collected according to the Chinese National Ethical Guidelines ('Code for Proper Secondary Use of Human Tissue', Chinese Federation of Medical Scientific Societies). Due to the limited available tumor material and follow-up information, only 44 of these osteosarcoma tumor samples were eligible for the present study. The mean age of the patients (mean ± SD) was 25.25±13.61 years (range 9–61; male/female ratio 32:12). Of the 44 eligible patients, 23 had non-metastatic tumors (53.3%). Osteoblastic osteosarcoma was the most common histopathological subtype and occurred in 29 patients (65.9%). The lower end of the femur was the most common site and was observed in 13 patients (29.5%). The patient distribution according to Rosen's histological grade (19,20) included stage I in 5 (11.3%), II in 16 (36.4%), III in 16 (36.4%) and stage IV in 7 patients (15.9%). Stages I and II were defined as low Rosen grade osteosarcoma specimens and stages III and IV were defined as high grade specimens. The patients were followed-up until death from the disease, or until their most recent clinical therapy at the end of the present study. The mean follow-up time (mean ± SD) was 4.26±1.99 years (range 0.5–9.0). All patients were diagnosed according to the osteosarcoma criteria defined by the World Health Organization. Written informed consent was obtained from each patient before he/she entered into the present study, and all study protocols were approved by the Ethics Committee for Clinical Research of Wuhan University, China.

The sections were cut from the formalin-fixed, paraffin-embedded osteosarcoma tissue and hydrated through graded alcohol solutions. For antigen unmasking, the sections were treated in a trypsin solution at 37°C for 10 min. The sections were then washed with deionized water and incubated with 3% H₂O₂ for 5 min. They were then incubated with the anti-IDH2 mAb at room temperature for 1 h, followed by a secondary antibody and the peroxidase-conjugated streptavidin-biotin complex (both from Santa Cruz Biotechnology, Santa Cruz, CA, USA) at 37°C for 30 min. The immunoreactivity was visualized with 3,3′-diaminobenzidine (DAB) (Zymed, South San Francisco, CA, USA). The negative controls were obtained by omitting the primary antibody.

IDH2 staining was detected in the mitochondria and was scored by adding the number of cells displaying clear tumor cell labeling; the intensity of staining was scored between 0 and 6 (32,33). The proportion score was as follows: 0 indicates negative staining; +1 indicates ≤25% positive labeling in tumor cells; +2 indicates 25–50% positive tumor cells; and +3 indicates >50% positive tumor cells. The intensity score was as follows: 0 indicates no staining; +1 indicates weak staining; +2 indicates intermediate staining; and +3 indicates strong staining (32,33). For statistical analysis, the osteosarcoma patients were grouped as either the low expression group (scored 1–4) or the high expression group (scored 5–6). At least 5 separated neoplastic infiltration foci were analyzed in each specimen, followed by the evaluation of 10 slides/patient and 6 sections/slide. The immunostaining was assessed by three independent observers. The slides were scanned using a microscope (Carl Zeiss AG, Germany) by reviewing the entire spot and the images were recorded using a digital camera (DC500; Leica) and Leica FW4000 software. The images were processed using Adobe Photoshop.

The Saos-2 and MG63 tumor cells (ATCC, LGC Promochem, Germany) were grown in Roswell Park Memorial Institute (RPMI)-1640 medium (Sigma-Aldrich, USA) with 10% fetal bovine serum (FBS) (Amresco, USA) and 0.1% penicillin/streptomycin and maintained in an atmosphere with 5% CO₂ at 37°C.

To downregulate IDH2, small interfering RNA (siRNA) sequences: LV-KD, 5′-GTGGACATCCAGCTAAAGTAT-3′, were inserted into the pLL3.7 shRNA lentiviral vector (Genesil, Wuhan, China). The pLL3.7 lentiviral vector LV-mIκB (Genesil), which was aimed at overexpressing the mutant IκB (mIκB), was constructed as previously described to suppress NF-κB activity (13). siRNA sequences for MMP-9 downregulation were: 5′-ACCACAACAUCACCUAUUGTT-3′ (34). The empty lentiviral vector (Genesil), LV-EV was used as a control. In some experiments, non-treated cells, named the NT cells, were used as another control. The lentiviral stocks were added to the Saos-2 and MG63 osteosarcoma cell lines.

The cells were infected with the lentivirus and selected with an 800 μg/ml G418 solution for the Saos-2 cells and a 500 μg/ml solution for the MG63 cells. The efficiency of the highest infection, as determined by G418 selection, was obtained at a multiplicity of infection (MOI) of 10 for the Saos-2 cells and 50 for the MG63 cells. The cells transfected with LV-KD or LV-EV were named the KD or EV cells, respectively. After selection, the efficiency of infection was verified by western blotting. Polyclonal populations and clones displaying higher levels of IDH2 downregulation were chosen for the subsequent studies. After IDH2 downregulation, LV-mIκB or LV-siMMP-9 was transfected into the cells for the NF-κB and MMP-9 assays, respectively.

The cell lysates were prepared using lysis buffer from the Dual-Luciferase assay kit (Promega, Madison, WI, USA) according to the manufacturer's instructions. The lysates were separated on 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gels and transferred onto polyvinylidene difluoride (PVDF) membranes (Millipore, Billerica, MA, USA). The membranes were then blocked with 5% skim milk in Tris-buffered saline with 0.05% Tween (TBST) and washed 6 times with TBST. The IDH2 and MMP-9 proteins were detected using rabbit polyclonal primary antibodies (Protein Technology Group, USA). NF-κB, Bcl-2, JNK, p-JNK, ERK, p-ERK, IκBα and p-IκBα were detected by mouse monoclonal primary antibodies (Santa Cruz Biotechnology). The β-actin proteins were recognized by a monoclonal β-actin-specific mouse IgG (Santa Cruz Biotechnology) and used as the internal loading control. The antibodies were diluted according to the manufacturer's instructions and incubated overnight at 4°C, followed by incubations with the peroxidase-conjugated goat anti-rabbit or anti-mouse immunoglobulins (1:2,000; Santa Cruz Biotechnology) in TBST for 1 h. The signals were developed using an enhanced chemiluminescent reagent (Pierce Biotechnology, Rockford, IL, USA).

A total of 3.5×10³ cells were seeded in each test well in a 96-well plate to detect cell growth. After 1–6 days of culture, the cells were washed with phosphate-buffered saline (PBS). MTT (5 mg/ml) was then added to each well (including the control) and the mixture was incubated at 37°C for 4 h. The culture medium was then replaced with an equal volume of dimethylsulfoxide (DMSO). After shaking the plate at room temperature for 10 min, the absorbance of each well was determined at 570 nm using a VersaMax microplate reader (Molecular Devices, Sunnyvale, CA, USA).

The cells were harvested by trypsinization, fixed with 70% pre-chilled alcohol and stored at 4°C overnight. The alcohol was then removed by centrifugation at 1,000 rpm for 5 min, and the cells were treated with 0.1% Triton-X and DNase-free RNase (10 mg/ml) for 30 min (35). The cell DNA was stained with 1 mg/ml propidium iodide (PI) for 15 min in the dark and analyzed using a flow cytometer (FACScan; Becton-Dickinson, New York, NY, USA). The relative proportions of cells in the G0–G1, S and G2-M phases of the cell cycle were determined from the flow cytometry data.

Cell invasion was determined using a two-chamber Transwell (Corning, New York, NY, USA). The upper surface of a polycarbonate membrane with 8-μm pores was coated with 1 mg/ml Matrigel (36). The cells (~10×10⁵) were suspended in RPMI-1640 serum-free medium (Gibco, USA) and placed in the upper chamber; RPMI-1640 medium containing 10% FBS was placed in the lower chamber. The cells were incubated at 37°C for 48 h with 5% CO₂ in an incubator. At the end of the incubation, the cells on the upper surface of the membrane were completely removed by wiping with a cotton swab. Then, the membrane was fixed with methanol and stained with 0.1% crystal violet. The cells that invaded the Matrigel and reached the lower surface of the membrane were counted and photographed under a microscope.

An NF-κB reporter plasmid, NF-κB-Luc, was constructed by cloning five repeats of the NF-κB regulatory elements into the pLuc plasmid (Stratagene, La Jolla, CA, USA) to drive luciferase expression. The empty plasmid, Luc, containing a minimal TATA box, was used as a negative control. Approximately 2.0×10⁶ cells were transiently transfected with 10 μg of the NF-κB reporter plasmid by electroporation (37). Then, the cells were seeded and incubated in 24-well plates at 37°C. After 24 h, the luciferase activity was analyzed using the Luciferase assay system (Promega) according to the manufacturer's instructions.

To obtain dissociated Saos2 and MG63 cells for the ROS assay, the culture medium was first removed and the cells were washed two times with RPMI-1640 serum-free medium (Sigma-Aldrich). DCFH-DA, diluted to a final concentration of 10 μM with RPMI-1640 medium, was added to the medium and incubated for 20 min at 37°C. The fluorescence was read at 488 nm for excitation and 530 nm for emission by flow cytometry (FACScan). The increase in the value compared to the control was viewed as the increase in intracellular ROS levels.

The IDH2 and/or MMP-9 down-regulated osteosarcoma cells were seeded in 6-well plates and incubated at 37°C. After 24 h, the medium was removed. Then, the cells were washed and incubated in serum-free medium for 48 h (37). The MMP-9 activity in the medium and cell lysate was detected using the Fluorokine E Human MMP-9 Activity assay kit (R&D Systems) according to the manufacturer's protocol.

The statistical analyses were performed using the SPSS 13.0 software package for Windows (SPSS, Inc., Chicago, IL, USA). Comparisons between groups were analyzed by the t-test or Mann-Whitney U test. Associations were assessed by Pearson's or Spearman's correlation coefficients. Event-free survival was calculated from the start of treatment to relapse or metastasis. Overall survival was calculated from the beginning of treatment to the last follow-up or death of the patient. The Kaplan-Meier method was used for the survival analysis. P<0.05 was considered to indicate a statistically significant result. P<0.01 was considered to indicate a highly statistically significant result.